PI(3,4,5)P3 and PI(4,5)P2 Lipids
Target Proteins with Polybasic
Clusters to the Plasma Membrane
Won Do Heo,1 Takanari Inoue,1 Wei Sun Park,1 Man Lyang Kim,1 Byung Ouk Park,2
Thomas J. Wandless,1 Tobias Meyer1*
Many signaling, cytoskeletal, and transport proteins have to be localized to the plasma membrane
(PM) in order to carry out their function. We surveyed PM-targeting mechanisms by imaging the
subcellular localization of 125 fluorescent protein–conjugated Ras, Rab, Arf, and Rho proteins.
Out of 48 proteins that were PM-localized, 37 contained clusters of positively charged amino
acids. To test whether these polybasic clusters bind negatively charged phosphatidylinositol
4,5-bisphosphate [PI(4,5)P2] lipids, we developed a chemical phosphatase activation method to
deplete PM PI(4,5)P2. Unexpectedly, proteins with polybasic clusters dissociated from the PM only
when both PI(4,5)P2 and phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P3] were depleted,
arguing that both lipid second messengers jointly regulate PM targeting.
mall guanosine triphosphatases (GTPases)
from the Ras, Rho, Arf, and Rab subfamilies often exert their role at the PM
where they control diverse signaling, cytoskeletal, and transport processes (1–3). KRas,
CDC42, and other family members require a
cluster of positively charged amino acids for PM
localization and activity (2, 4). In vitro studies
indicate that the physiological PM binding partner of such polybasic clusters could be phosphatidylserine, which has one negative charge, or the
less abundant lipid second messenger PI(4,5)P2,
which has four negative charges (5–7). We took
a genomic survey approach and investigated
PM-targeting mechanisms by confocal imaging
of 125 cyan fluorescent protein (CFP)–tagged
constitutively active small GTPases (8). Expression in NIH3T3 and HeLa cells showed that 48
small GTPases were fully or partially localized to
the PM (Fig. 1A and fig. S1).
Thirty-seven of these PM-localized small
GTPases had C-terminal polybasic clusters
consisting of four or more Lys or Arg residues
at positions 5 to 20 from the C terminus (Fig. 1B
and fig. S1). Polybasic clusters were found in
three forms: They were present together with
N-terminal myristoylation consensus sequences
(as in Arl4) (9) or with C-terminal prenylation
consensus sequences (as in KRas) (5, 6, 10), or
they lacked lipid modifications (as in Rit) (11).
We called these three combinations polybasicmyristoyl, polybasic-prenyl, and polybasicnonlipid PM-targeting motifs, respectively. A
number of remaining PM-targeted small GTPases
had a combined prenylation and palmitoylation
consensus sequence that mediated PM targeting
S
1
Department of Molecular Pharmacology, 318 Campus
Drive, Clark Building, Stanford University Medical School,
Stanford, CA 94305, USA. 2Division of Applied Life Science
(BK21 Program) and Environmental Biotechnology National Core Research Center, Gyeongsang National University, Jinju 660-701, Korea.
*To whom correspondence should be addressed. E-mail:
tobias1@stanford.edu
1458
without requiring polybasic amino acids (as does
that of HRas) (Fig. 1, A and C) (5, 12). Arf6
lacked a specific targeting motif and was only
localized to the PM in its guanosine triphosphate
(GTP)–bound form (fig. S2) (13). The sequence
homology comparison of PM-localized small
GTPases in Fig. 1C shows that closely homologous small GTPases can have different targeting
motifs. We also confirmed, for the examples of
Rit and KRas, that polybasic targeting motifs
alone can be sufficient for PM targeting (Fig. 1D).
To test whether polybasic clusters are anchored to the PM by binding to PI(4,5)P2 (14),
we hydrolyzed PM PI(4,5)P2 by rapid targeting
of Inp54p, a 5′ specific PI(4,5)P2 phosphatase
(15), to the PM. This method is based on a
PM-localized FK506-binding protein (FKBP12)–
rapamycin–binding (FRB) construct and a cytosolic Inp54p enzyme conjugated with FKBP12
(CF-Inp) that can be translocated to the PM by
chemical heterodimerization by using a rapamycin
analog, iRap (16).
In experiments where we monitored PI(4,5)P2
using a yellow fluorescent protein (YFP)–
conjugated pleckstrin homology (PH) domain
from phospholipase Cd (PLCd) (17), PM translocation of CF-Inp triggered a rapid and near
complete dissociation of the YFP-PLCd-PH domain from the PM (Fig. 2A). Despite this marked
reduction in PI(4,5)P2 concentration, only a small
fraction of the polybasic-nonlipid tail fragment of
Rit dissociated from the PM (Fig. 2B), which
suggests that PI(4,5)P2 is not alone responsible
for PM targeting. Parallel experiments suggested
that phosphatidylinositol 3,4,5-trisphosphate
[PI(3,4,5)P3] might be involved, because stimulation of cells with platelet-derived growth factor
Fig. 1. A survey of the subcellular localization of 125 small GTPases shows that most PM-localized
small GTPases have targeting motifs with clusters of polybasic amino acids. (A) Confocal images of
the subcellular localization of CFP-conjugated small GTPases in NIH3T3 cells (full set of images in
NIH3T3 and HeLa cells in fig. S1). The four main PM-targeting motifs are represented in the image
panels. (B) Correlation between PM localization and the presence of lysine residues in a region 5 to
20 amino acids from the C terminus. (C) Phylogenetic tree of 48 small GTPases that were identified
to be partially or fully localized to the PM. Individual membrane targeting elements are color
coded: red for polybasic clusters and blue, green, and orange for palmitoyl, prenyl, and myristoyl
consensus sequences, respectively. (D) Twenty-amino-acid-long C-terminal tail fragments of Rit and
KRas are PM-localized. Lack of PM targeting of a KRas tail fragment without the polybasic region
(right). Scale bars, 10 mm.
1 DECEMBER 2006
VOL 314
SCIENCE
www.sciencemag.org
Downloaded from www.sciencemag.org on January 15, 2007
REPORTS
(PDGF) led to a small increase in PM localization
of the Rit tail and this small effect could be
reversed by addition of an inhibitor of phosphoinositide 3-kinase (PI 3-kinase), LY294002
(LY29) (Fig. 2C; control experiments in fig. S3).
Strikingly, the combined reduction of both
PI(4,5)P2 and PI(3,4,5)P3 concentration triggered a dissociation of most Rit tail protein from
Fig. 2. Depletion of PI(4,5)P2 and PI(3,4,5)P3 dissociates Rit, Rin, and MARCKS ED with polybasicnonlipid targeting motifs from the PM. (A) Development of a chemically inducible translocation
method to deplete PI(4,5)P2 from the inner leaflet of the PM (22, 23). The PI(4,5)P2 biosensor YFPPLCd-PH was cotransfected to monitor depletion of PI(4,5)P2. (B) Depletion of PI(4,5)P2 caused
only a small reduction in the PM localization of YFP-conjugated Rit tail. (C) A small, but significant,
PDGF receptor–mediated increase in Rit tail PM localization can be reversed by addition of the PI3kinase inhibitor LY29 (before and 9 min after addition of 5 mM iRap). The bar graph shows a
quantification of the same experiment. The PM dissociation index is the relative ratio of internal
over PM fluorescence; that is, F1cyt/F1PM * F0PM/F0cyt, with F0 and F1 as the fluorescent intensities
before and after PI(4,5)P2 and PI(3,4,5)P3 depletion. (D) Joint reduction in PI(4,5)P2 and PI(3,4,5)P3
triggered a near-complete dissociation of Rit tail and MARCKS ED from the PM. (E) Quantitative
analysis of the CF-Inp and/or LY29-triggered PM dissociation of MARCKS ED, and Rit and Rin tails.
PLCd-PH and HRas tails are shown as controls. The inactive LY29 analog LY30 was used as a control
(24). Scale bars, 10 mm.
Fig. 3. Polybasic subclusters and hydrophobic amino acids are required for PM targeting by polybasicnonlipid targeting motifs. (A) Statistical analysis of the relative sequence location of positively charged
amino acids in nonprenylated small GTPases. (B) Loss of PM targeting of Rit tail fragment after removal
of a single subcluster with positively charged amino acids. (C) Identification of a tryptophan residue in
the polybasic regions of Rit and Rin that mediates PM over nuclear localization. Full-length small
GTPases, as well as tail fragment mutants, are shown. (D) Identification of two additional hydrophobic
amino acid residues that contribute to the PM targeting of Rit. Scale bars, 10 mm.
www.sciencemag.org
SCIENCE
VOL 314
the PM (Fig. 2, D and E). The effect of reducing
either PI(4,5)P2 or PI(3,4,5)P3 concentration
alone on Rit tail localization was relatively small
(Fig. 2E).
We also found that both PI(4,5)P2 and
PI(3,4,5)P3 have to be lowered to significantly
dissociate a tail fragment of Rin and a polybasic
effector domain peptide from myristoylated alaninerich C kinase substrate (MARCKS) protein
(MARCKS ED) from the PM (Fig. 2, D and E).
This MARCKS ED peptide was included because it has been extensively used for in vitro
biochemical studies of the interaction between
polybasic amino acids and phosphatidylserine
and PI(4,5)P2 (7, 14, 18, 19). HRas, which has a
prenyl-palmitoyl PM-targeting motif without a
polybasic cluster, did not dissociate from the
PM after depletion of PI(4,5)P2 and PI(3,4,5)P3
(Fig. 2E). Control experiments with the same constructs in HeLa cells showed similar results (fig.
S4), and an analysis of the dissociation kinetics
showed that PM dissociation occurs within
minutes after CF-Inp activation and LY29 addition (fig. S5A). We also verified that PM dissociation of polybasic proteins occurred when we
combined PI(4,5)P2 depletion with addition of
wortmannin, an alternative PI 3-kinase inhibitor,
or with expression of a dominant negative PI 3kinase inhibitory construct (20) (figs. S6 and S7).
Additional control experiments are shown in figs.
S8 to S12, including in vitro lipid-blot assays that
showed enhanced affinity of proteins with polybasic clusters for PI(3,4,5)P3 over PI(4,5)P2.
These control experiments strengthen the argument that PI(4,5)P2 and the less abundant
PI(3,4,5)P3 both serve as PM anchors for proteins with polybasic clusters.
Insights into the electrostatic binding mechanism between positively charged polybasic
clusters and negatively charged polyphosphoinositides came from a sequence comparison of
nonprenylated PM-targeted small GTPases. Most
of these proteins contain two or three subclusters
of polybasic residues in the C-terminal tail. Each
subcluster spans about four or five amino acids,
and the mean distance between subclusters is
nine amino acids (Fig. 3A). Consistent with a
need for multiple subclusters, the removal of a
flanking subcluster in the Rit tail was sufficient to
abolish PM targeting (Rit tail 199 to 219) (Fig.
3B). This suggests that PM targeting results from
additive binding energy of individual subclusters
that each electrostatically interact with a PI(4,5)P2
or PI(3,4,5)P3 lipid. Selectivity for polyphosphoinositides over phosphatidylserine may occur
because of opposing high relative-charge densities of polyphosphoinositides and polybasic
subclusters (7).
We then investigated differences between the
targeting mechanisms of the three polybasicnonlipid, polybasic-myristoyl, and polybasicprenyl PM-targeting motifs. A distinct feature
of the polybasic-nonlipid targeting motifs is
their similarity to nuclear localization sequences.
We found hydrophobic amino acids to be im-
1 DECEMBER 2006
Downloaded from www.sciencemag.org on January 15, 2007
REPORTS
1459
Fig. 4. Depletion of PI(4,5)P2 and
PI(3,4,5)P3 dissociates proteins with
polybasic-myristoyl and polybasicprenyl targeting motifs from the
PM. (A) Aligned sequences of the
N terminus (6 amino acids) and C
terminus (20 amino acids) of six
ARF family members. Confocal images of an Arl7 mutant construct
lacking the N-terminal myristoylation motif (left), a C-terminal tagged
Arl7 control construct (middle), and
a mutant construct with an internal
CFP tag and a flexible C-terminal
polybasic Arl7 tail (right). (B) PM localization motifs in geranylgeranylated
Rab35. Confocal images from left to
right: localization of wild-type Rab35,
wild-type tail fragment, wild-type tail
lacking geranylgeranylation motif
(∆CC), Rab35 tail ∆CC mutant with
KRas caax (Rab35 tail caax), and
Rab35 tail caax mutant lacking polybasic amino acids (Rab35 tail caax
∆pB). (C) Depletion of PI(4,5)P2 by
CF-Inp activation without L29 addition causes only a minimal PM dissociation of Arl7, KRas tail fragment, and Rab35. (D) Quantitative analysis of the much larger PM dissociation of
polybasic-lipid modified proteins after depletion of PI(4,5)P2 and PI(3,4,5)P3 by CF-Inp activation and addition of LY29. Scale bars, 10 mm.
portant for selective PM localization, because
site-directed mutagenesis of a single hydrophobic
amino acid (Trp204Ala in Rit or Trp202Ala in Rin)
led to a complete loss of PM targeting and a
strong nuclear localization (Fig. 3C). A loss of
PM targeting but with less nuclear targeting
could also be observed for Leu207Ala and
Phe211Ala Rit mutants (Fig. 3D) and for hydrophobic amino acid mutants of the small GTPases
GEM and RAD (fig. S13). Thus, hydrophobic
amino acids strengthen PM binding of polybasicnonlipid motifs and prevent the polybasic cluster from functioning as a nuclear localization
sequence.
The polybasic-myristoyl PM-targeting motifs
have the distinct feature of a separated N-terminal
myristoylation consensus sequence and a Cterminal polybasic cluster. Mutant constructs
showed that effective PM targeting of Arl7 required an N-terminal myristoyl motif (left panel,
Fig. 4A), as well as a flexible C-terminal polybasic tail (middle versus right panel, Fig. 4A),
which suggests that the two ends of the protein
synergistically support PM targeting.
The polybasic-prenyl PM-targeting motif
includes proteins such as KRas for which a 20–
amino acid tail sequence is sufficient for farnesylation and PM targeting (Fig. 1D), as well as Rab35
for which an intact GTPase domain is required
for geranylgeranylation (21) and for PM targeting (second and third panels, Fig. 4B). We
further compared the roles of farnesylation and
geranylgeranylation by creating a Rab35 mutant
with a consensus CAAX farnesylation sequence
in place of the geranylgeranylation sequence.
This mutant showed PM localization indistinguishable from that of the geranylgeranylated
1460
Rab35 (fourth panel, Fig. 4B). We also confirmed
that the PM targeting of RAB35 requires a
polybasic cluster (last panel, Fig. 4B). This
shows that the polybasic-geranylgeranyl motifs
of Rab35 can be equally effective in PM targeting
as the polybasic-farnesyl motif of KRas, which
supports the notion that both types of prenylation
motifs can be grouped into a single polybasicprenyl PM-targeting motif.
We then tested whether PI(4,5)P 2 and
PI(3,4,5)P3 also regulate polybasic-myristoyl and
polybasic-prenyl targeting motifs. As for the
polybasic-nonlipid targeting motif, depletion of
PI(4,5)P2 alone triggered only a minor reduction
in PM localization of the Arl7 polybasicmyristoyl targeting motif and the KRas and
Rab35 polybasic-prenyl motifs (Fig. 4C). Depletion of both PI(4,5)P2 and PI(3,4,5)P3 triggered
significant PM dissociation of all polybasicmyristoyl and polybasic-prenyl constructs tested
(Fig. 4D) with a kinetics similar to that of Rit
(fig. S5), which suggests that PI(4,5)P2 and
PI(3,4,5)P3 have the same role for PM localization for all three types of polybasic PM-targeting
motifs. HRas was again included as a control
protein without a polybasic cluster.
Our study shows that polybasic PM-targeting
motifs are built from two parts, an unspecific
membrane-targeting part that can be hydrophobic amino acids, myristoyl groups, or prenyl
groups and a polybasic targeting part that
provides PM specificity by binding of positively charged amino acid clusters to negatively charged PI(4,5)P2 and PI(3,4,5)P3 lipids in
the PM. This gives PI(4,5)P2 and PI(3,4,5)P3 a
ubiquitous role in regulating signaling, cytoskeletal, and transport proteins and argues that
1 DECEMBER 2006
VOL 314
SCIENCE
these lipid second messengers function as
signaling hubs in cellular control systems.
References and Notes
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
M. Fivaz, T. Meyer, Neuron 40, 319 (2003).
M. N. Teruel, T. Meyer, Cell 103, 181 (2000).
J. L. Guan, Science 303, 773 (2004).
D. Michaelson et al., J. Cell Biol. 152, 111 (2001).
K. A. Cadwallader, H. Paterson, S. G. Macdonald,
J. F. Hancock, Mol. Cell. Biol. 14, 4722 (1994).
F. Ghomashchi, X. Zhang, L. Liu, M. H. Gelb, Biochemistry
34, 11910 (1995).
S. McLaughlin, D. Murray, Nature 438, 605 (2005).
W. D. Heo, T. Meyer, Cell 113, 315 (2003).
A. Schurmann et al., J. Biol. Chem. 269, 15683 (1994).
E. Choy et al., Cell 98, 69 (1999).
C. H. Lee, N. G. Della, C. E. Chew, D. J. Zack, J. Neurosci.
16, 6784 (1996).
O. Rocks et al., Science 307, 1746 (2005).
H. Radhakrishna, J. G. Donaldson, J. Cell Biol. 139, 49
(1997).
S. McLaughlin, J. Wang, A. Gambhir, D. Murray, Annu.
Rev. Biophys. Biomol. Struct. 31, 151 (2002).
T. Inoue, W. D. Heo, J. S. Grimley, T. J. Wandless, T. Meyer,
Nat. Methods 2, 415 (2005).
D. Raucher et al., Cell 100, 221 (2000).
T. P. Stauffer, S. Ahn, T. Meyer, Curr. Biol. 8, 343 (1998).
J. Wang, A. Arbuzova, G. Hangyas-Mihalyne, S. McLaughlin,
J. Biol. Chem. 276, 5012 (2001).
C. Chapline, K. Ramsay, T. Klauck, S. Jaken, J. Biol. Chem.
268, 6858 (1993).
S. Poser, S. Impey, K. Trinh, Z. Xia, D. R. Storm, EMBO J.
19, 4955 (2000).
M. C. Seabra, C. Wasmeier, Curr. Opin. Cell Biol. 16, 451
(2004).
B.-C. Suh, T. Inoue, T. Meyer, B. Hille, Science 314,
1454 (2006); published online 21 September 2006;
10.1126/science.1131163.
Correspondence about the iRap inducible enzyme systems
should be directed to T. Inoue (e-mail: jctinoue@
stanford.edu).
T. W. Poh, S. Pervaiz, Cancer Res. 65, 6264 (2005).
We thank A. Salmeen, M. F. Teruel, M. Fivaz, and
other members of the Meyer laboratory for support;
A. R. Koh and S. H. Ryu (POSTECH) for critical reading the
www.sciencemag.org
Downloaded from www.sciencemag.org on January 15, 2007
REPORTS
REPORTS
This work was supported by grants from National Institute
of Mental Health and National Institute of General
Medical Sciences, NIH, to T.M.
28 August 2006; accepted 11 October 2006
Published online 9 November 2006;
10.1126/science.1134389
Include this information when citing this paper.
Supporting Online Material
www.sciencemag.org/cgi/content/full/1134389/DC1
Materials and Methods
A Genome-Wide Association Study
Identifies IL23R as an Inflammatory
Bowel Disease Gene
Richard H. Duerr,1,2 Kent D. Taylor,3,4 Steven R. Brant,5,6 John D. Rioux,7,8 Mark S. Silverberg,9
Mark J. Daly,8,10 A. Hillary Steinhart,9 Clara Abraham,11 Miguel Regueiro,1 Anne Griffiths,12
Themistocles Dassopoulos,5 Alain Bitton,13 Huiying Yang,3,4 Stephan Targan,4,14
Lisa Wu Datta,5 Emily O. Kistner,15 L. Philip Schumm,15 Annette T. Lee,16 Peter K. Gregersen,16
M. Michael Barmada,2 Jerome I. Rotter,3,4 Dan L. Nicolae,11,17 Judy H. Cho18*
The inflammatory bowel diseases Crohn’s disease and ulcerative colitis are common, chronic
disorders that cause abdominal pain, diarrhea, and gastrointestinal bleeding. To identify genetic
factors that might contribute to these disorders, we performed a genome-wide association study.
We found a highly significant association between Crohn’s disease and the IL23R gene on
chromosome 1p31, which encodes a subunit of the receptor for the proinflammatory cytokine
interleukin-23. An uncommon coding variant (rs11209026, c.1142G>A, p.Arg381Gln) confers
strong protection against Crohn’s disease, and additional noncoding IL23R variants are
independently associated. Replication studies confirmed IL23R associations in independent cohorts
of patients with Crohn’s disease or ulcerative colitis. These results and previous studies on the
proinflammatory role of IL-23 prioritize this signaling pathway as a therapeutic target in
inflammatory bowel disease.
rohn’s disease (CD) and ulcerative colitis (UC), the two common forms of idiopathic inflammatory bowel disease (IBD),
are chronic, relapsing inflammatory disorders of
the gastrointestinal tract. Each has a peak age of
onset in the second to fourth decades of life and
prevalences in European ancestry populations
that average about 100 to 150 per 100,000 (1, 2).
Although the precise etiology of IBD remains to
be elucidated, a widely accepted hypothesis is
that ubiquitous, commensal intestinal bacteria
trigger an inappropriate, overactive, and ongoing mucosal immune response that mediates
intestinal tissue damage in genetically susceptible individuals (1). Genetic factors play an important role in IBD pathogenesis, as evidenced
by the increased rates of IBD in Ashkenazi Jews,
familial aggregation of IBD, and increased
concordance for IBD in monozygotic compared
to dizygotic twin pairs (3). Moreover, genetic
analyses have linked IBD to specific genetic
variants, especially CARD15 variants on chromosome 16q12 and the IBD5 haplotype (spanning the organic cation transporters, SLC22A4
and SLC22A5, and other genes) on chromosome
5q31 (3–7). CD and UC are thought to be related
disorders that share some genetic susceptibility
loci but differ at others.
The replicated associations between CD
and variants in CARD15 and the IBD5 haplotype do not fully explain the genetic risk for
C
CD, so we performed a genome-wide association study testing 308,332 autosomal single
nucleotide polymorphisms (SNPs) on the Illumina HumanHap300 Genotyping BeadChip (8).
Our study population consisted of 567 nonJewish, European ancestry patients with ileal CD
and 571 non-Jewish controls. We initially
focused on ileal CD, the most common location
of CD, to minimize pathogenic heterogeneity.
After exclusion of study subjects with genotype
completion rates less than 94%, we included 547
cases and 548 controls in subsequent analyses
(8). Single-marker allelic tests were performed
using c2 statistics for all autosomal markers.
Three SNPs had nearly two orders of magnitude
greater significance compared to the next most
significant markers, and they are the only
markers that remain significant at the 0.05 level
after Bonferroni correction. Two of the three
markers, rs2066843 (P = 2.86 × 10−9, corrected
P = 8.82 × 10−4) and rs2076756 (P = 5.12 × 10−10,
corrected P = 1.58 × 10−4), are in the known CD
susceptibility gene, CARD15 (4, 5). The third
marker, rs11209026 (P = 5.05 × 10−9, corrected
P = 1.56 × 10−3), is a nonsynonymous SNP
(c.1142G>A, p.Arg381Gln) in the IL23R gene
(GenBank accession: NM_144701, GeneID:
149233) on chromosome 1p31. This gene encodes a subunit of the receptor for the proinflammatory cytokine, interleukin-23 (IL-23),
and is therefore an intriguing functional can-
www.sciencemag.org
SCIENCE
VOL 314
Figs. S1 to S13
References
didate. In addition to Arg381Gln, nine other
markers in IL23R and in the intergenic region
between IL23R and the adjacent IL-12 receptor,
beta-2 gene (IL12RB2), had association Pvalues < 0.0001 in the non-Jewish, ileal CD
case-control cohort (Table 1 and table S1a).
We next tested for association of IL23R
markers in an independent ileal CD case-control
cohort, consisting of 401 patients and 433 controls, all of Jewish ancestry (8). Significant associations were observed for several of the same
markers that were associated in the non-Jewish
cohort (Table 1 and table S1b). In a combined
analysis of the data from the two ileal CD casecontrol cohorts (8), nine markers had highly
significant association P-values ranging from
1.60 × 10−9 to 3.36 × 10−13 (Table 1 and table S1b).
We then extended the replication study by
performing family-based association testing of
27 IL23R region markers in an independent cohort of 883 nuclear families in which both par1
Division of Gastroenterology, Hepatology and Nutrition,
Department of Medicine, School of Medicine, University of
Pittsburgh, University of Pittsburgh Medical Center Presbyterian, Mezzanine Level, C-Wing, 200 Lothrop Street, Pittsburgh,
PA 15213, USA. 2Department of Human Genetics, Graduate
School of Public Health, University of Pittsburgh, Crabtree
A300, 130 Desoto Street, Pittsburgh, PA 15261, USA. 3Medical
Genetics Institute, Cedars-Sinai Medical Center, 8700 Beverly
Boulevard, Los Angeles, CA 90048, USA. 4IBD Center, Division
of Gastroenterology, Cedars-Sinai Medical Center, 8700
Beverly Boulevard, Los Angeles, CA 90048, USA. 5Harvey M.
and Lyn P. Meyerhoff Inflammatory Bowel Disease Center,
Department of Medicine, Johns Hopkins University School of
Medicine, B136, 1503 East Jefferson Street, Baltimore, MD
21231, USA. 6Department of Epidemiology, Bloomberg School
of Public Health, Johns Hopkins University, 615 North Wolfe
Street, Baltimore, MD 21205, USA. 7Université de Montréal
and the Montreal Heart Institute, S-6400, 5000 Belanger
Street, Montreal, Quebec H1T 1C8, Canada. 8Medical and
Population Genetics Program, Broad Institute of MIT and
Harvard, 7 Cambridge Center, Cambridge, MA 02142, USA.
9
Mount Sinai Hospital IBD Centre, University of Toronto, 441–
600 University Avenue, Toronto, Ontario M5G 1X5, Canada.
10
Massachusetts General Hospital, Harvard Medical School,
185 Cambridge Street, Boston, MA 02114, USA. 11Department
of Medicine, University of Chicago, 5841 South Maryland
Avenue, Chicago, IL 60637, USA. 12Department of Pediatrics,
The Hospital for Sick Children, 555 University Avenue, Toronto,
Ontario M5G 1X8, Canada. 13Royal Victoria Hospital, McGill
University Health Centre, 687 Pine Avenue West, Montreal,
Quebec H3A 1A1, Canada. 14Immunobiology Research
Institute, Cedars-Sinai Medical Center, Davis 4063, 8700
Beverly Boulevard, Los Angeles, CA 90048, USA. 15Department
of Health Studies, University of Chicago, 5841 South Maryland
Avenue, Chicago, IL 60637, USA. 16The Feinstein Institute
for Medical Research, 350 Community Drive, Manhasset, NY
11030, USA. 17Department of Statistics, University of Chicago,
5734 South University Avenue, Chicago, IL 60637, USA. 18IBD
Center, Section of Digestive Diseases, Departments of Medicine
and Genetics, Yale University, S155A, 300 Cedar Street, New
Haven, CT 06519, USA.
Downloaded from www.sciencemag.org on January 15, 2007
manuscript; S. McLaughlin (SUNY Stonybrook) and B. Hille
(U. Washington) for discussions; and James Whalen for
assisting with the in vitro lipid-binding assays. T.I. is a
recipient of a fellowship from the Quantitative Chemical
Biology Program. B.O.P. was supported by a grant from
KOSEF/MOST EB-NCRC (grant no. R15-2003-012-01001-0)
and by scholarships from the Brain Korea 21 program.
*To whom correspondence should be addressed. E-mail:
judy.cho@yale.edu
1 DECEMBER 2006
1461